[libcxx-commits] [libcxx] [libc++] Tiny optimizations for is_permutation (PR #129565)

[Imad Aldij via libcxx-commits]

imdj wrote:

>  Indeed, this is the speedup I get when I drop `std::ref` from the `mismatch` call and ensure that we vectorize on AppleClang

I'm missing some piece of the puzzle. I couldn't reach those numbers. Like mentioned above in the benchmarks , I'm currently getting:
```
OVERALL_GEOMEAN                                                                   -0.0022         -0.0022             0             0             0             0
```

Just dropping `std::ref()` like such `auto __result = std::mismatch(__first1, __last1, __first2, __pred);` will lead to copies/failed tests and even then, the performance still took a hit:
```
OVERALL_GEOMEAN                                                                   -0.0013         -0.0013             0             0             0             0
```

I also tried using `std::__mismatch` directly to avoid making copies something like this:
```cpp
__identity __ident;
auto __result = std::__mismatch(__first1, __last1, __first2, __pred, __ident, __ident);
```

but that also ended up with worse results
```
OVERALL_GEOMEAN                                                                   -0.0009         -0.0009             0             0             0             0
```
What am I doing wrong :(

https://github.com/llvm/llvm-project/pull/129565